Academic FAQ Chatbot using Retrieval Augmented Generation

Abstract

Traditional methods of accessing information from academic guideline documents often lead to inefficient manual searches and repetitive queries.  This research addresses these challenges by developing an academic Frequently Asked Question (FAQ) chatbot using the Retrieval Augmented Generation (RAG) approach. The study aims to simplify academic information access on academic documents through an AI-integrated chatbot.  The methodology includes document collection, data preprocessing, and vector representation via embedding techniques. A retrieval system is built using a vector database for similarity-based searches, integrated with a Large Language Model (LLM).  This research used All-MiniLM-L6-v2 model for embeddings, while Gemini-2.0-flash is utilized as the LLM for response generation.  The chatbot interface is developed using Flask framework.  User surveys indicate higher satisfaction with the developed RAG-based chatbot compared to traditional services. This research contributes significantly by demonstrating an effective, adaptive, and cost-efficient AI solution that enhances academic information services.